Convolutional Neural Networks¶
Project: Write an Algorithm for a Dog Identification App¶
In this notebook, some template code has already been provided for you, and you will need to implement additional functionality to successfully complete this project. You will not need to modify the included code beyond what is requested. Sections that begin with '(IMPLEMENTATION)' in the header indicate that the following block of code will require additional functionality which you must provide. Instructions will be provided for each section, and the specifics of the implementation are marked in the code block with a 'TODO' statement. Please be sure to read the instructions carefully!
Note: Once you have completed all of the code implementations, you need to finalize your work by exporting the Jupyter Notebook as an HTML document. Before exporting the notebook to html, all of the code cells need to have been run so that reviewers can see the final implementation and output. You can then export the notebook by using the menu above and navigating to File -> Download as -> HTML (.html). Include the finished document along with this notebook as your submission.
In addition to implementing code, there will be questions that you must answer which relate to the project and your implementation. Each section where you will answer a question is preceded by a 'Question X' header. Carefully read each question and provide thorough answers in the following text boxes that begin with 'Answer:'. Your project submission will be evaluated based on your answers to each of the questions and the implementation you provide.
Note: Code and Markdown cells can be executed using the Shift + Enter keyboard shortcut. Markdown cells can be edited by double-clicking the cell to enter edit mode.
The rubric contains optional "Stand Out Suggestions" for enhancing the project beyond the minimum requirements. If you decide to pursue the "Stand Out Suggestions", you should include the code in this Jupyter notebook.
Why We're Here¶
In this notebook, you will make the first steps towards developing an algorithm that could be used as part of a mobile or web app. At the end of this project, your code will accept any user-supplied image as input. If a dog is detected in the image, it will provide an estimate of the dog's breed. If a human is detected, it will provide an estimate of the dog breed that is most resembling. The image below displays potential sample output of your finished project (... but we expect that each student's algorithm will behave differently!).
In this real-world setting, you will need to piece together a series of models to perform different tasks; for instance, the algorithm that detects humans in an image will be different from the CNN that infers dog breed. There are many points of possible failure, and no perfect algorithm exists. Your imperfect solution will nonetheless create a fun user experience!
The Road Ahead¶
We break the notebook into separate steps. Feel free to use the links below to navigate the notebook.
- Step 0: Import Datasets
- Step 1: Detect Humans
- Step 2: Detect Dogs
- Step 3: Create a CNN to Classify Dog Breeds (from Scratch)
- Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)
- Step 5: Write your Algorithm
- Step 6: Test Your Algorithm
Step 0: Import Datasets¶
Make sure that you've downloaded the required human and dog datasets:
Download the dog dataset. Unzip the folder and place it in this project's home directory, at the location
/dog_images.Download the human dataset. Unzip the folder and place it in the home directory, at location
/lfw.
Note: If you are using a Windows machine, you are encouraged to use 7zip to extract the folder.
In the code cell below, we save the file paths for both the human (LFW) dataset and dog dataset in the numpy arrays human_files and dog_files.
import numpy as np
from glob import glob
# load filenames for human and dog images
dog_files = np.array(glob('./dog_images/*/*/*'))
human_files = np.array(glob('./human_images/*/*'))
# print number of images in each dataset
print('There are %d total human images' % len(dog_files))
print('There are %d total dog images' % len(human_files))
Step 1: Detect Humans¶
In this section, we use OpenCV's implementation of Haar feature-based cascade classifiers to detect human faces in images.
OpenCV provides many pre-trained face detectors, stored as XML files on github. We have downloaded one of these detectors and stored it in the haarcascades directory. In the next code cell, we demonstrate how to use this detector to find human faces in a sample image.
import cv2
import matplotlib.pyplot as plt
%matplotlib inline
# extract pre-trained face detector
face_cascade = cv2.CascadeClassifier('./haarcascades/haarcascade_frontalface_alt.xml')
# load color (BGR) image
img = cv2.imread(human_files[0])
# convert BGR image to grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# find faces in image
faces = face_cascade.detectMultiScale(gray)
# print number of faces detected in the image
print('Number of faces detected:', len(faces))
# get bounding box for each detected face
for (x, y, w, h) in faces:
# add bounding box to color image
cv2.rectangle(img, (x, y), (x+w, y+h), (255, 0, 0), 2)
# convert BGR image to RGB for plotting
cv_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
# display the image, along with bounding box
plt.imshow(cv_rgb)
plt.show()
Before using any of the face detectors, it is standard procedure to convert the images to grayscale. The detectMultiScale function executes the classifier stored in face_cascade and takes the grayscale image as a parameter.
In the above code, faces is a numpy array of detected faces, where each row corresponds to a detected face. Each detected face is a 1D array with four entries that specifies the bounding box of the detected face. The first two entries in the array (extracted in the above code as x and y) specify the horizontal and vertical positions of the top left corner of the bounding box. The last two entries in the array (extracted here as w and h) specify the width and height of the box.
Write a Human Face Detector¶
We can use this procedure to write a function that returns True if a human face is detected in an image and False otherwise. This function, aptly named face_detector, takes a string-valued file path to an image as input and appears in the code block below.
# returns "True" if face is detected in image stored at img_path
def face_detector(img_path):
img = cv2.imread(img_path)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
faces = face_cascade.detectMultiScale(gray)
return len(faces) > 0
(IMPLEMENTATION) Assess the Human Face Detector¶
Question 1: Use the code cell below to test the performance of the face_detector function.
- What percentage of the first 100 images in
human_fileshave a detected human face? - What percentage of the first 100 images in
dog_fileshave a detected human face?
Ideally, we would like 100% of human images with a detected face and 0% of dog images with a detected face. You will see that our algorithm falls short of this goal, but still gives acceptable performance. We extract the file paths for the first 100 images from each of the datasets and store them in the numpy arrays human_files_short and dog_files_short.
Answer: (You can print out your results and/or write your percentages in this cell)
from tqdm import tqdm
human_files_short = human_files[:100]
dog_files_short = dog_files[:100]
#-#-# Do NOT modify the code above this line. #-#-#
## TODO: Test the performance of the face_detector algorithm
## on the images in human_files_short and dog_files_short.
human_detected = 0.0
dog_detected = 0.0
num_files = len(human_files_short)
for i in range(0, num_files):
human_path = human_files_short[i]
dog_path = dog_files_short[i]
if face_detector(human_path) == True:
human_detected += 1
if face_detector(dog_path) == True:
dog_detected += 1
print('Haar Face Detection')
print('The percentage of the detected face - Humans:{0:.0%}'.format(human_detected / num_files))
print('The percentage of the detected face - Dogs:{0:.0%}'.format(dog_detected / num_files))
We suggest the face detector from OpenCV as a potential way to detect human images in your algorithm, but you are free to explore other approaches, especially approaches that make use of deep learning :). Please use the code cell below to design and test your own face detection algorithm. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.
Step 2: Detect Dogs¶
In this section, we use a pre-trained model to detect dogs in images.
Obtain Pre-trained VGG-16 Model¶
The code cell below downloads the VGG-16 model, along with weights that have been trained on ImageNet, a very large, very popular dataset used for image classification and other vision tasks. ImageNet contains over 10 million URLs, each linking to an image containing an object from one of 1000 categories.
import torch
import torchvision.models as models
# define VGG16 model
VGG16 = models.vgg16(pretrained=True)
Given an image, this pre-trained VGG-16 model returns a prediction (derived from the 1000 possible categories in ImageNet) for the object that is contained in the image.
(IMPLEMENTATION) Making Predictions with a Pre-trained Model¶
In the next code cell, you will write a function that accepts a path to an image (such as 'dogImages/train/001.Affenpinscher/Affenpinscher_00001.jpg') as input and returns the index corresponding to the ImageNet class that is predicted by the pre-trained VGG-16 model. The output should always be an integer between 0 and 999, inclusive.
Before writing the function, make sure that you take the time to learn how to appropriately pre-process tensors for pre-trained models in the PyTorch documentation.
from PIL import Image
import torchvision.transforms as transforms
from torch.autograd import Variable
human_files_short = human_files[:100]
dog_files_short = dog_files[:100]
def VGG16_predict(img_path):
'''
Use pre-trained VGG-16 model to obtain index corresponding to
predicted ImageNet class for image at specified path
Args:
img_path: path to an image
Returns:
Index corresponding to VGG-16 model's prediction
'''
## TODO: Complete the function.
## Load and pre-process an image from the given img_path
img = Image.open(img_path)
# VGG-16 takes 224x224 images as input, resize!
# Convert PIL image into Tersor
# Normailize input images to make its elements from 0 to 1
data_transform = transforms.Compose([transforms.RandomResizedCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])])
# Apply the transformation above
img = data_transform(img)
# PyTorch pretrained models expect the Tensor dims to be (num input imgs, num color channels, height, width).
# Currently however, we have (num color channels, height, width); let's fix this by inserting a new axis.
# Insert the new axis at index 0 i.e. in front of the other axes/dims.
img = img.unsqueeze(0)
# Now that we have preprocessed our img, we need to convert it into a
# Variable; PyTorch models expect inputs to be Variables. A PyTorch Variable is a
# wrapper around a PyTorch Tensor.
img = Variable(img)
# Returns a Tensor of shape (batch, num class labels)
prediction = VGG16(img)
prediction = prediction.data.numpy().argmax()
## Return the *index* of the predicted class for that image
return prediction
(IMPLEMENTATION) Write a Dog Detector¶
While looking at the dictionary, you will notice that the categories corresponding to dogs appear in an uninterrupted sequence and correspond to dictionary keys 151-268, inclusive, to include all categories from 'Chihuahua' to 'Mexican hairless'. Thus, in order to check to see if an image is predicted to contain a dog by the pre-trained VGG-16 model, we need only check if the pre-trained model predicts an index between 151 and 268 (inclusive).
Use these ideas to complete the dog_detector function below, which returns True if a dog is detected in an image (and False if not).
### returns "True" if a dog is detected in the image stored at img_path
def dog_detector(img_path):
class_index = VGG16_predict(img_path)
if class_index >= 151 and class_index <= 268:
return True
else:
return False
(IMPLEMENTATION) Assess the Dog Detector¶
Question 2: Use the code cell below to test the performance of your dog_detector function.
- What percentage of the images in
human_files_shorthave a detected dog? - What percentage of the images in
dog_files_shorthave a detected dog?
Answer: (You can print out your results and/or write your percentages in this cell)
### TODO: Test the performance of the dog_detector function
### on the images in human_files_short and dog_files_short.
human_detected = 0.0
dog_detected = 0.0
num_files = len(human_files_short)
for i in range(0, num_files):
human_path = human_files_short[i]
dog_path = dog_files_short[i]
if dog_detector(human_path) == True:
human_detected += 1
if dog_detector(dog_path) == True:
dog_detected += 1
print('VGG-16 Prediction')
print('The percentage of the detected dog - Humans: {0:.0%}'.format(human_detected / num_files))
print('The percentage of the detected dog - Dogs: {0:.0%}'.format(dog_detected / num_files))
We suggest VGG-16 as a potential network to detect dog images in your algorithm, but you are free to explore other pre-trained networks (such as Inception-v3, ResNet-50, etc). Please use the code cell below to test other pre-trained PyTorch models. If you decide to pursue this optional task, report performance on human_files_short and dog_files_short.
Step 3: Create a CNN to Classify Dog Breeds (from Scratch)¶
Now that we have functions for detecting humans and dogs in images, we need a way to predict breed from images. In this step, you will create a CNN that classifies dog breeds. You must create your CNN from scratch (so, you can't use transfer learning yet!), and you must attain a test accuracy of at least 10%. In Step 4 of this notebook, you will have the opportunity to use transfer learning to create a CNN that attains greatly improved accuracy.
We mention that the task of assigning breed to dogs from images is considered exceptionally challenging. To see why, consider that even a human would have trouble distinguishing between a Brittany and a Welsh Springer Spaniel.
| Brittany | Welsh Springer Spaniel |
|---|---|
 |
 |
It is not difficult to find other dog breed pairs with minimal inter-class variation (for instance, Curly-Coated Retrievers and American Water Spaniels).
| Curly-Coated Retriever | American Water Spaniel |
|---|---|
 |
 |
Likewise, recall that labradors come in yellow, chocolate, and black. Your vision-based algorithm will have to conquer this high intra-class variation to determine how to classify all of these different shades as the same breed.
| Yellow Labrador | Chocolate Labrador | Black Labrador |
|---|---|---|
 |
 |
 |
We also mention that random chance presents an exceptionally low bar: setting aside the fact that the classes are slightly imabalanced, a random guess will provide a correct answer roughly 1 in 133 times, which corresponds to an accuracy of less than 1%.
Remember that the practice is far ahead of the theory in deep learning. Experiment with many different architectures, and trust your intuition. And, of course, have fun!
(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset¶
Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dog_images/train, dog_images/valid, and dog_images/test, respectively). You may find this documentation on custom datasets to be a useful resource. If you are interested in augmenting your training and/or validation data, check out the wide variety of transforms!
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
%matplotlib inline
# Check if CUDA is available
use_cuda = torch.cuda.is_available()
print('CUDA 사용가능:', use_cuda)
It had not been easy to get above the 10 % of accuracy, so, as the mentor told me, I decide to imitate VGG-16 model. Below is the transformation code for train, valid and test sets.
import torchvision.transforms as transforms
# Declare the transforms for train, valid and test sets.
# Imitate the VGG-16 model.
# Resize images because the input size of VGG-16 is 224x224
# Convert to Tensor
# Normalize images because the values of images should be loaded between [0 - 1]
transforms = {
# Use RandomHorizontalFlip() to augement data in the train transformation
'train' : transforms.Compose([transforms.Resize(256),
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])]),
'valid' : transforms.Compose([transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])]),
'test' : transforms.Compose([transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])])
}
Create loaders for each dataset.
from torchvision import datasets
from torchvision import utils
import os
# Number of subprocesses, if it's 0, it uses the main process.
num_workers = 0
# How many samples will be loaded for one batch?
batch_size = 20
# Create image datasets (train, valid, test)
image_datasets = {x: datasets.ImageFolder(os.path.join('dog_images', x), transforms[x])
for x in ['train', 'valid', 'test']}
# Create data loaders (train, valid, test)
data_loaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=batch_size,
shuffle=True, num_workers=num_workers)
for x in ['train', 'valid', 'test']}
# Decrease batch size because of the out of memory in the GPU Instance
test_loader = torch.utils.data.DataLoader(image_datasets['test'], shuffle=True,
batch_size=15)
The codes below display the simple information of datasets.
# Check the dataset sizes
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'valid', 'test']}
print('Number of records of training dataset: {}'.format(dataset_sizes['train']))
print('Number of records of validation dataset: {}'.format(dataset_sizes['valid']))
print('Number of records of test dataset: {}'.format(dataset_sizes['test']))
# Get the all the breed labels
class_names = image_datasets['train'].classes
print(class_names)
# Get the number of classes
n_classes = len(class_names)
print('Number of classes: {}'.format(n_classes))
# Display one record (input, label)
# The image should be normalized, the label is a integer value between 0 - 132
data_loaders['train'].dataset[6679]
Visualize Image Transformation¶
To check if the transformations are applied correctly, visualize the image samples.
def visualize_sample_images(inp):
inp = inp.numpy().transpose((1, 2, 0))
inp = np.clip(inp, 0, 1)
fig = plt.figure(figsize=(50, 25))
plt.axis('off')
plt.imshow(inp)
plt.pause(0.001)
# Get a batch of training data.
inputs, classes = next(iter(data_loaders['train']))
# Convert the batch to a grid.
grid = utils.make_grid(inputs)
# Display!
visualize_sample_images(grid)
Question 3: Describe your chosen procedure for preprocessing the data.
- How does your code resize the images (by cropping, stretching, etc)? What size did you pick for the input tensor, and why?
- Did you decide to augment the dataset? If so, how (through translations, flips, rotations, etc)? If not, why not?
Answer:
I chose the size
(244, 244) usingtransforms.resize()function. Most of the images are too big. If you use it in the orignal size of images, the calculation would take so much time and memories.Yes, I augmented the dataset. The 8351 dog images, it seems small dataset to me. And I didn't want predictions to be changed depends on the size, rotation or translations of images. I used
RandomHorizontalFlip()method.
(IMPLEMENTATION) Model Architecture¶
Create a CNN to classify dog breed. Use the template in the code cell below.
The image below is the architecture of VGG-16.
import torch.nn as nn
# define the CNN architecture
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
## Follow the architecture of VGG-16
# Size 224
self.conv1_1 = nn.Conv2d(3, 64, kernel_size=3, padding=1)
self.conv1_2 = nn.Conv2d(64, 64, kernel_size=3, padding=1)
# Size 112
self.conv2_1 = nn.Conv2d(64, 128, kernel_size=3, padding=1)
self.conv2_2 = nn.Conv2d(128, 128, kernel_size=3, padding=1)
# Size 56
self.conv3_1 = nn.Conv2d(128, 256, kernel_size=3, padding=1)
self.conv3_2 = nn.Conv2d(256, 256, kernel_size=3, padding=1)
self.conv3_3 = nn.Conv2d(256, 256, kernel_size=3, padding=1)
# Size 28
self.conv4_1 = nn.Conv2d(256, 512, kernel_size=3, padding=1)
self.conv4_2 = nn.Conv2d(512, 512, kernel_size=3, padding=1)
self.conv4_3 = nn.Conv2d(512, 512, kernel_size=3, padding=1)
# Size 14
self.conv5_1 = nn.Conv2d(512, 512, kernel_size=3, padding=1)
self.conv5_2 = nn.Conv2d(512, 512, kernel_size=3, padding=1)
self.conv5_3 = nn.Conv2d(512, 512, kernel_size=3, padding=1)
# Batch Normalization is a technique to improve performance and stability
# of an artifical neural network.
# It provides zero mean and unit variance as inputs to any layers.
self.batch_norm64 = nn.BatchNorm2d(64)
self.batch_norm128 = nn.BatchNorm2d(128)
self.batch_norm256 = nn.BatchNorm2d(256)
self.batch_norm512 = nn.BatchNorm2d(512)
self.max_pool = nn.MaxPool2d(kernel_size=2, stride=2)
self.relu = nn.ReLU(inplace=True)
self.dropout = nn.Dropout()
# Size 7
self.fc1 = nn.Linear(512 * 7 * 7, 4096)
self.fc2 = nn.Linear(4096, 4096)
# The Last fully connected layer's output is 133(Number of breeds)
self.fc3 = nn.Linear(4096, 133)
# Feedforward
def forward(self, x):
x = self.relu(self.batch_norm64(self.conv1_1(x)))
x = self.relu(self.batch_norm64(self.conv1_2(x)))
x = self.max_pool(x)
x = self.relu(self.batch_norm128(self.conv2_1(x)))
x = self.relu(self.batch_norm128(self.conv2_2(x)))
x = self.max_pool(x)
x = self.relu(self.batch_norm256(self.conv3_1(x)))
x = self.relu(self.batch_norm256(self.conv3_2(x)))
x = self.relu(self.batch_norm256(self.conv3_3(x)))
x = self.max_pool(x)
x = self.relu(self.batch_norm512(self.conv4_1(x)))
x = self.relu(self.batch_norm512(self.conv4_2(x)))
x = self.relu(self.batch_norm512(self.conv4_3(x)))
x = self.max_pool(x)
x = self.relu(self.batch_norm512(self.conv5_1(x)))
x = self.relu(self.batch_norm512(self.conv5_2(x)))
x = self.relu(self.batch_norm512(self.conv5_3(x)))
x = self.max_pool(x)
# It returns a new tensor which has a different size
# and it's the same data of self tensor
# The -1 means inferring the size from other dimensions.
x = x.view(x.size(0), -1)
x = self.dropout(self.relu(self.fc1(x)))
x = self.dropout(self.relu(self.fc2(x)))
x = self.fc3(x)
return x
# Create CNN instance!
model_scratch = Net()
# If CUDA is avaliable, Move Tensors to GPU
if use_cuda:
model_scratch.cuda()
Question 4: Outline the steps you took to get to your final CNN architecture and your reasoning at each step.
Answer: First, it was really hard for me to get an accuracy more than 10%. So, as one of the Udacity mentors said, I tried to imitate the VGG16 model.
outline:
Input: a fixed size, 224x224 RGB image
Kernel Size: 3x3, the smallest size to capture the notion of left/right, up/down, center)
Padding: it is 1 for 3x3 kernel, to keep the same spatial resolution
MaxPooling: 2x2 with stride of 2 pixels, to reduce the size of image and the amount of parameters in half and to capture the most useful pixels(computation reduced!)
Activation Function: ReLU, quick to evaluate, it does not saturate (if the input is very high or very low, the gradient is very, very small)
Batch Normalization 2D: It is a technique to provide any laer in a Neural Network with inputs that are zero mean or unit variance
Convolutional Layers -> (Input channels, Output channels):
1-1 (3, 64) 1-2 (64, 64)
2-1 (64, 128) 2-2 (128, 128)
3-1 (128, 256) 3-2 (256, 256) 3-3 (256, 256)
4-1 (256, 512) 4-2 (512, 512) 4-3 (512, 512)
5-1 (512, 512) 5-2 (512, 512) 5-3 (512, 512)
Fully Connected Layers: two fully connected layers with 4096 units(with dropout) each and one with 133 units (representing the dog breed classes, without dropout)
(IMPLEMENTATION) Specify Loss Function and Optimizer¶
Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_scratch, and the optimizer as optimizer_scratch below.
import torch.optim as optim
### TODO: select loss function
criterion_scratch = nn.CrossEntropyLoss()
### TODO: select optimizer
optimizer_scratch = optim.SGD(model_scratch.parameters(), lr=0.001, momentum=0.9)
(IMPLEMENTATION) Train and Validate the Model¶
Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_scratch.pt'.
def train(n_epochs, train_loader, valid_loader,
model, optimizer, criterion, use_cuda, save_path):
"""returns trained model"""
# initialize tracker for minimum validation loss
valid_loss_min = np.Inf
for epoch in range(1, n_epochs+1):
# initialize variables to monitor training and validation loss
train_loss = 0.0
valid_loss = 0.0
###################
# train the model #
###################
for batch_idx, (data, target) in enumerate(train_loader):
# move to GPU
if use_cuda:
data, target = data.cuda(), target.cuda()
## find the loss and update the model parameters accordingly
## record the average training loss, using something like
## train_loss = train_loss + ((1 / (batch_idx + 1)) * (loss.data - train_loss))
# clear the gradients of all optimized variables
optimizer.zero_grad()
# forward pass
output = model(data)
# calculate batch loss
loss = criterion(output, target)
# backward pass
loss.backward()
# parameter update
optimizer.step()
# update training loss
train_loss += loss.item() * data.size(0)
######################
# validate the model #
######################
for batch_idx, (data, target) in enumerate(valid_loader):
# move to GPU
if use_cuda:
data, target = data.cuda(), target.cuda()
## update the average validation loss
# forward pass
output = model(data)
# batch loss
loss = criterion(output, target)
# update validation loss
valid_loss += loss.item() * data.size(0)
# calculate average losses
train_loss = train_loss / len(train_loader.dataset)
valid_loss = valid_loss / len(valid_loader.dataset)
# print training/validation statistics
print('Epoch: {}\tTraining Loss: {:.6f}\t Validation Loss: {:.6f}'.
format(epoch, train_loss, valid_loss))
## TODO: save the model if validation loss has decreased
if valid_loss <= valid_loss_min:
print('Validation loss decreased ({:.6f} --> {:.6f}). Saving model...'.
format(valid_loss_min, valid_loss))
torch.save(model.state_dict(), save_path)
valid_loss_min = valid_loss
# return trained model
return model
n_epochs = 20
# train the model
model_scratch = train(n_epochs, data_loaders['train'], data_loaders['valid'], model_scratch,
optimizer_scratch, criterion_scratch, use_cuda, 'model_scratch.pt')
# load the model that got the best validation accuracy
model_scratch.load_state_dict(torch.load('model_scratch.pt'))
Test the Model¶
Let's apply the test data to the trained model. The codes below calculate the test loss and its accuracy. Our goal is to get above 10 % of the accuracy. 강아지 이미지의 테스트 데이터셋을 학습시킨 모델에 적용시켜봅니다. 아래의 코드는 test loss와 accuracy를 계산해줍니다. 10% 이상의 test accuracy를 가지는게 목표입니다!
def test(loader, model, criterion, use_cuda):
# monitor test loss and accuracy
test_loss = 0.
correct = 0.
total = 0.
for batch_idx, (data, target) in enumerate(loader):
# move to GPU
if use_cuda:
data, target = data.cuda(), target.cuda()
# forward pass: compute predicted outputs by passing inputs to the model
output = model(data)
# calculate the loss
loss = criterion(output, target)
# update average test loss
test_loss = test_loss + ((1 / (batch_idx + 1)) * (loss.data - test_loss))
# convert output probabilities to predicted class
pred = output.data.max(1, keepdim=True)[1]
# compare predictions to true label
correct += np.sum(np.squeeze(pred.eq(target.data.view_as(pred))).cpu().numpy())
total += data.size(0)
print('Test Loss: {:.6f}\n'.format(test_loss))
print('\nTest Accuracy: %2d%% (%2d/%2d)' % (100. * correct / total, correct, total))
# call test function
test(test_loader, model_scratch, criterion_scratch, use_cuda)
Step 4: Create a CNN to Classify Dog Breeds (using Transfer Learning)¶
You will now use transfer learning to create a CNN that can identify dog breed from images. Your CNN must attain at least 60% accuracy on the test set.
(IMPLEMENTATION) Specify Data Loaders for the Dog Dataset¶
Use the code cell below to write three separate data loaders for the training, validation, and test datasets of dog images (located at dogImages/train, dogImages/valid, and dogImages/test, respectively).
If you like, you are welcome to use the same data loaders from the previous step, when you created a CNN from scratch.
(IMPLEMENTATION) Model Architecture¶
Use transfer learning to create a CNN to classify dog breed. Use the code cell below, and save your initialized model as the variable model_transfer.
# Load VGG-16 model
model_transfer = models.vgg16(pretrained=True)
# Freeze the pre-trained weights
for param in model_transfer.features.parameters():
param.required_grad = False
# Get the input of the last layer of VGG-16
n_inputs = model_transfer.classifier[6].in_features
# Create a new layer(n_inputs -> 133)
# The new layer's requires_grad will be automatically True.
last_layer = nn.Linear(n_inputs, 133)
# Change the last layer to the new layer.
model_transfer.classifier[6] = last_layer
# Print the model.
print(model_transfer)
if use_cuda:
model_transfer = model_transfer.cuda()
Question 5: Outline the steps you took to get to your final CNN architecture and your reasoning at each step. Describe why you think the architecture is suitable for the current problem.
Answer: I chose the VGG16 model. I've already imitated it from scratch at the step 3. But I wondered the differences between my model and the original VGG16 model. And I also wondered how the accuracy will go from the pre-trained model, because I had limited computing power, dataset and time to train.
i thought the VGG16 is sutable for the current problem. Because it already trained large dataset. So I initialized randomly the wieght in the new fully connected layer, and the rest of the weights using the pre-trained weights. And overfitting is not as much of a concern when training on a large data set. And the model classifies like my problem needs.
(IMPLEMENTATION) Specify Loss Function and Optimizer¶
Use the next code cell to specify a loss function and optimizer. Save the chosen loss function as criterion_transfer, and the optimizer as optimizer_transfer below.
criterion_transfer = nn.CrossEntropyLoss()
optimizer_transfer = optim.SGD(model_transfer.classifier.parameters(), lr=0.001)
(IMPLEMENTATION) Train and Validate the Model¶
Train and validate your model in the code cell below. Save the final model parameters at filepath 'model_transfer.pt'.
n_epochs = 20
model_transfer = train(n_epochs, data_loaders['train'], data_loaders['valid'], model_transfer,
optimizer_transfer, criterion_transfer, use_cuda, 'model_transfer.pt')
# load the model that got the best validation accuracy (uncomment the line below)
model_transfer.load_state_dict(torch.load('model_transfer.pt'))
(IMPLEMENTATION) Test the Model¶
Try out your model on the test dataset of dog images. Use the code cell below to calculate and print the test loss and accuracy. Ensure that your test accuracy is greater than 60%.
test(data_loaders['test'], model_transfer, criterion_transfer, use_cuda)
(IMPLEMENTATION) Predict Dog Breed with the Model¶
Write a function that takes an image path as input and returns the dog breed (Affenpinscher, Afghan hound, etc) that is predicted by your model.
# CPU or GPU
device = torch.device("cuda:0" if use_cuda else "cpu")
# Load the trained model 'model_transfer.pt'
model_transfer.load_state_dict(torch.load('model_transfer.pt', map_location='cpu'))
import torchvision.transforms as transforms
def predict_breed_transfer(img_path):
# class names without number (Affenpinscher, Brussels griffon...)
# class names with number (001.Affenpinscher', '038.Brussels_griffon')
class_names_without_number = [item[4:].replace("_", " ") for item in image_datasets['train'].classes]
class_names_with_number = image_datasets['train'].classes
# Load image
img = Image.open(img_path)
# Image Preprocessing
transform_predict = transforms.Compose([transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])])
# Get Tensor
img_tensor = transform_predict(img)
# without the code below, it occurs an error !
# RuntimeError: expected stride to be a single integer value or a list of 1 values
# to match the convolution dimensions, but got stride=[1, 1]
img_tensor = img_tensor.unsqueeze_(0)
# Send the tensor to the device(GPU or CPU)
img_tensor = img_tensor.to(device)
# PyTorch's models need inputs as a form of Variable.
# Variable is a wrapper of Pytorch Tensor.
img_var = Variable(img_tensor)
# Get output
output = model_transfer(img_var)
# Get the probability of breeds
softmax = nn.Softmax(dim=1)
preds = softmax(output)
# Get three breeds which has the highest probabilities.
top_preds = torch.topk(preds, 3)
# Get the names of breed for displaying
labels_without_number = [class_names_without_number[i] for i in top_preds[1][0]]
labels_with_number = [class_names_with_number[i] for i in top_preds[1][0]]
# Get the probabilities as a form of Tensor
probs = top_preds[0][0]
return labels_without_number, labels_with_number, probs
Step 5: Write your Algorithm¶
Write an algorithm that accepts a file path to an image and first determines whether the image contains a human, dog, or neither. Then,
- if a dog is detected in the image, return the predicted breed.
- if a human is detected in the image, return the resembling dog breed.
- if neither is detected in the image, provide output that indicates an error.
You are welcome to write your own functions for detecting humans and dogs in images, but feel free to use the face_detector and human_detector functions developed above. You are required to use your CNN from Step 4 to predict dog breed.
Some sample output for our algorithm is provided below, but feel free to design your own user experience!
(IMPLEMENTATION) Write your Algorithm¶
import glob
# Load test images
human_files = glob.glob('./test_images/human_images/*')
dog_files = glob.glob('./test_images/dog_images/*')
# Display input image
def display_image(img_path):
# Display image
img = Image.open(img_path)
_, ax = plt.subplots()
ax.imshow(img)
plt.axis('off')
plt.show()
# Display dog breed images
def display_breeds(labels):
fig = plt.figure(figsize=(16,4))
for i, label in enumerate(labels):
subdir = ''.join(['dog_images/valid/', label + '/'])
file = random.choice(os.listdir(subdir))
path = ''.join([subdir, file])
img = Image.open(path)
ax = fig.add_subplot(1,3,i+1)
ax.imshow(img, cmap="gray", interpolation='nearest')
plt.title(label.split('.')[1])
plt.axis('off')
plt.show()
import random
def run_app(img_path):
# Get the probabilities and the labels.
labels_without_number, labels_with_number, probs = predict_breed_transfer(img_path)
# If it's a dog,
if probs[0] > 0.3:
# Display the input image
print("It's a dog!")
display_image(img_path)
# Display the predicted breeds and its probablities
print("Predicted breeds and its probabilities:\n")
sentence = ""
for pred_label, prob in zip(labels_without_number, probs):
print(pred_label)
print('{:.2f}%'.format(100*prob))
print('\n')
# Display predicted breed images
display_breeds(labels_with_number)
# If it's a human,
elif face_detector(img_path):
# Display the input image
print("It's a human!")
display_image(img_path)
# Display the most resembled breeds and its probabilities
print("Resembled breeds and its probabilities:\n")
for pred_label, prob in zip(labels_without_number, probs):
print(pred_label)
print('{:.2f}%'.format(100*prob))
print('\n')
# Display the resembled breeds
display_breeds(labels_with_number)
# Not a dog and not a human,
else:
# Display the input image
print("I can't detect if it's a human or dog!")
display_image(img_path)
print('\n')
print('\n')
Step 6: Test Your Algorithm¶
In this section, you will take your new algorithm for a spin! What kind of dog does the algorithm think that you look like? If you have a dog, does it predict your dog's breed accurately? If you have a cat, does it mistakenly think that your cat is a dog?
(IMPLEMENTATION) Test Your Algorithm on Sample Images!¶
Test your algorithm at least six images on your computer. Feel free to use any images you like. Use at least two human and two dog images.
Question 6: Is the output better than you expected :) ? Or worse :( ? Provide at least three possible points of improvement for your algorithm.
Answer:
- Kante, a football player was classified as a dog. It needs to modify the threshold.(Done!)
- Some same codes are repeated, so it needs to be modularized. (Done!)
- To compare performance among models, run_app() should take model as a input parameter.
# Test!
for file in np.hstack((human_files[:5], dog_files[:5])):
run_app(file)
